EC 571 Advanced Econometrics

Homework 3

Part I

Based on a classical study of investment demand (Greene [2008], pp.250-252, Grunfeld and Griliches [1960], Boot and deWitt [1960]), the model is:

I_it = α_it + β_itF_it + γ_itC_it + ε_it

where
i = 10 firms: GM, CH, GE, WE, US, AF, DM, GY, UN, IBM.
t = 20 years: 1935-1954.
I_it = Gross investment.
F_it = Market value.
C_it = Value of the stock of plant and equipment.
e_it = Error term.

Data of above 3 variables for 10 companies are available:

General Motors (GM)	Chrysler (CH)
General Electric (GE)	Westinghouse (WE)
U. S. Steel (US)	Atlantic Refining (AF)
Diamond Match (DM)	Goodyear (GY)
Union Oil (UN)	IBM

Formulate the panel data model as:

I_it = α_i + γ_t + βF_it + γC_it + ε_it

That is, with common slopes, only the intercept parameters vary across firms and over time. To study the individual (i.e., firm) effects alone, we further assume γ_t = 0. Similarly, to study the time effects alone, we assume α_i = 0.

1. Estimate the firm effects panel data model, considering both fixed effects and randowm effects. Which model should be used based on hypothesis testings?

2. Estimate the time effects panel data model, considering both fixed effects and randowm effects. Which model should be used based on hypothesis testings?

3. Can we estimate both firm effects and time effects (combined two-way effects) panel data model? Explain and interpret the best model you selected.

Part II

• Part 1:
  1. Column 1 = Person id (ranging from 1 to 2178),
  2. Column 2 = Education,
  3. Column 3 = Log of hourly wage,
  4. Column 4 = Potential experience,
  5. Column 5 = Time trend.
• Part 2:
  1. Column 1 = Time invariant ability,
  2. Column 2 = Mother's education,
  3. Column 3 = Father's education,
  4. Column 4 = Dummy variable for residence in a broken home,
  5. Column 5 = Number of siblings.

The regression model under consideration is

ln Wageit =	αi + β1 Educationit + β2 Expeienceit + β3 Experienceit2
	+ β4 Broken_Homei + β5 Siblingsi + εit

Using all of the 17,919 observations of the data provided, estimate (1) fixed effects panel data model; (2) random effects panel data model; and (3) compare and test to select either fixed effects or random effects model.